PowrMatch is a complete two‑sided marketplace for Australia's energy sector — connecting specialist consultants with the organisations driving the energy transition, and carrying every engagement through contracts, timesheets and settlement. Designed, built and shipped in five months.
The Challenge
Australia's energy transition runs on specialists — grid engineers, market analysts, regulatory advisors, hydrogen and storage strategists. Most work independently, and the organisations that need them still find them the slow way: through personal networks, recruiters and long email threads.
The harder half comes after the handshake. A single engagement generates a contract to negotiate and sign, weekly timesheets to submit and approve, invoices to raise with GST, a platform fee to account for, and a payment to reconcile. In most of the market that lives across PDFs, inboxes and spreadsheets, with no shared record of truth — and no way for either side to see where things actually stand.
PowrMatch was built to hold that entire lifecycle in one system, and to hold it for both sides at once — because a consultant and a client are looking at the same engagement from opposite ends, and both need it to be right.
What We Built
Every stage below is live in the platform. The sequence is genuinely dependent — each step produces the artifact the next one needs, so nothing falls into a gap between systems.
Separate guided journeys for consultants and clients, with LinkedIn sign‑in or email. Every account passes an admin approval gate before going live, and Australian Business Numbers are verified against the national register. Clients can post a requirement before they even have an account — the draft is held and migrated the moment they verify.
A multi‑criteria matching engine evaluates consultants against a requirement across eight dimensions — work type, energy area, role, rate, experience, availability, skills and location. Clients also search the pool directly, with results carrying real engagement history: completed jobs, reviews, live availability.
Clients send interview requests; consultants accept or decline. Accepted requests hand straight off to Calendly for time selection, with the booked slot flowing back automatically and calendar invitations issued to both parties. A reconciliation service independently verifies bookings, so a scheduled meeting still lands even when a webhook does not arrive.
On hire, the platform generates a complete contract from the agreed engagement terms and opens a two‑party signing envelope. Both signatures are tracked in real time; when the second lands, the contract locks and the engagement activates automatically. Every signature is recorded with a full audit trail.
Weekly timesheets arrive pre‑filled and ready to adjust, with a review step before submission. Clients approve or return them with a reason. The system understands engagement state — it will not accept time against work that has already closed — and sends automatic reminders when a week goes unsubmitted.
Approved timesheets roll straight into a formatted invoice — hours and rates itemised, GST calculated, platform fee disclosed in full. No re‑keying, no separate accounting step, and no ambiguity about what is being charged or why.
Two settlement routes, both fully tracked: card payment through Stripe Connect, or bank transfer with confirmation flowing to the consultant automatically. The platform fee is retained and the balance is transferred to the consultant's connected account — so money moves without anyone chasing it.
An engagement cannot simply be declared finished — the platform validates that every invoice is paid and every timesheet approved before it will close. Dashboards flag engagements approaching their end date well in advance, and extension is a first‑class path: new dates, new rate, revised scope, no re‑contracting from scratch.
Under the Hood
A marketplace that handles contracts and payments has to be right the first time. These are the engineering decisions that made that possible.
Database schema, API contracts and UI components are defined once and shared across the entire stack. A change to a data model surfaces immediately as a compile error in every screen that touches it — entire categories of production bug are eliminated before code ever runs.
Six external services — payments, scheduling, e‑signature, email, identity and business verification — each with webhook handling, retry logic and reconciliation. Where a provider cannot guarantee delivery, the platform verifies independently rather than assuming.
Meeting times, timesheet weeks and engagement dates are handled in the viewer's own zone throughout — stored absolutely, rendered locally, and always labelled. A client in Perth and a consultant in Brisbane see the same appointment at the right hour.
Thirteen event types drive a dual‑channel notification system — an in‑app panel that refreshes continuously, plus branded email. Scheduled jobs chase unsubmitted timesheets and flag engagements nearing their end date without anyone remembering to look.
Consultants, clients and administrators each get a purpose‑built dashboard over the same data, with role‑based access enforced at the API. Admins get verification queues, payment tracking and platform analytics without a separate back office.
The platform ships with a seventeen‑document technical reference covering architecture, every feature area, the full API surface and the complete data model — so the system is maintainable by any competent team, not only the one that built it.
How We Built It
A platform of this scope would traditionally take a team the better part of a year. It was delivered in five months and 667 commits — because AI was used as genuine engineering capacity, not as a novelty.
Every AI‑produced change in this project arrived with a specific file, a specific line, and a stated reason. Nothing was accepted on confidence alone. That single discipline is what separates AI‑assisted engineering from AI‑generated code — and it is why the result is a system that can be audited, documented and handed over.
Authentication, dual onboarding, dashboards, search, interview flows and notifications — over half the project's total commits landed in the first month, with an AI agent working against a continuously maintained specification.
Contracts, e‑signature, timesheets, invoicing and Stripe. Velocity was deliberately traded for care as the system took on financial consequences — every path reviewed rather than generated.
Once the platform worked, AI was pointed at it in reverse — reading the source to produce the full technical reference set and a systematic 59‑point engineering review across eleven quality dimensions.
The hardest defects — a cross‑timezone display fault and a calendar sync edge case — were traced to root cause in written analysis before a line changed, then fixed once and correctly.
The advantage isn't typing speed. It's holding a 14,000‑line module and six integrations in view at the same time.
Engineering Rigour
Most delivery teams ship and move on. On PowrMatch, the finished platform was put through a systematic engineering review of its own source — fifty‑nine assessment points across eleven dimensions, covering security, authorisation, concurrency, error handling, performance, accessibility and long‑term maintainability.
Each point was recorded with a precise location, a severity, and a recommended remedy, then tracked as a prioritised backlog rather than a one‑off report. It is the difference between believing a system is sound and being able to demonstrate it — and it means the roadmap beyond launch is driven by evidence instead of guesswork.
What's Next
The platform's foundation — structured profiles, structured requirements, and a complete history of real engagements — is exactly the substrate machine learning needs. This is the roadmap it unlocks.
Move beyond exact‑criteria matching so that a consultant describing “NEM settlements” is surfaced for a requirement asking for “wholesale market operations”. Meaning‑based search over profiles and requirements, learning from which matches actually convert to engagements.
Guide clients to complete, well‑specified briefs at the moment of writing — the single highest‑leverage improvement available, because match quality is capped by requirement quality.
Turn consultant onboarding from a long multi‑step form into an upload and a confirmation, extracting skills, roles and experience automatically — directly attacking the biggest source of signup drop‑off.
Rank verification, invoice and payout queues by risk and summarise each item, so administrative review scales with the marketplace instead of constraining it.
Got a project? A wild idea? Or just want to say hey?
We're here for all of it — reach out anytime.
